The headline here under-serves the article in my opinion: this is a fascinating, deep explanation of how the memory market works and why increased demand for HBM (used by big GPU racks) hurts the availability of wafers for DDR and LPDDR (used by laptops and phones).
> these memory makers have learned a very particular lesson from the unforgiving history [deep drops in demand] of their industry: always leave demand unmet
I can't blame them for keeping some reserve demand ready so they keep having customers over the years.
It's more than that though. Memory is a series of boom-and-bust cycles. Also, it costs around $15 billion and takes 2-3 years to build a fab. So if you were to start now, you'd have a pretty good chance of hitting the next bust right after spending $15 billion. It's not necessarily some sinister conspiracy, it's because execs are reluctant to be the ones to take the blame for setting fire to $15 billion. Looks kinda bad on your CV.
I can't say I've noticed specifically. I have two tracfone accounts and a Cricket account so used to use the Android phone with Cricket after two months - a free phone. But tracfone was bought by Verizon and they being them immediately changed the unlock period from two months to a year. So that to me kills my use case of free smartphones as I don't want to spend any money on phones.
The deep dive on memory market dynamics and the LLM bubble distortion is great. But another cause of declining smart phone sales is simply that the devices have matured and aren't improving at nearly the same rate.
From 2008 to around 2015, upgrading every two years could make a meaningful difference. From ~2015 to ~2020 upgrading every three years might be worth considering. I just upgraded my top of the line flagship after nearly six years. And I actually looked for compelling reasons to buy a new phone every year since 2023. There just weren't any.
Frankly, this latest flagship phone is pretty underwhelming. It's slightly faster at a few things. The battery lasts a little longer. The screen can get a little brighter. The camera is supposed to a little better. But those are just the claimed improvements. I haven't actually noticed any of them in daily use because they weren't issues with my 2020 flagship phone either. Otherwise, the new phone is almost exactly the same size, same weight, same resolution, same look and same capabilities. I only upgraded because I was long out of contract and it was a only a couple hundred bucks for a $1400 MSRP phone with a new contract and a trade-in of the old phone.
Yeah I'm still using an iPhone 12 from late 2020 and it's honestly still fine in terms of its processing power, cameras, features etc. for anything I use it for. It does need a new battery but even that has degraded way slower than earlier phones I've had.
I was using an iPad Pro from late 2018 (mostly just for casual web browsing, reading documents and watching video on, I still do all my real work on laptop/desktops) as well until this year, and would have kept using it if I hadn't accidentally dropped it in water. I don't really notice much difference at all between my old one (when it worked) and the new iPad Air I replaced it with, except for the battery being a little better and having a bit more ram being nice (websites in background tabs are less likely to be purged from memory when I come back to them).
Yeah, I mostly upgraded to get the magnetic charger on the back of my iPhone. Weren't a lot of other compelling reasons - the already pretty great cameras improved a bit, the screen is a bit nicer, but otherwise its much the same slab of glass.
The MacBook Pro on which I’m writing this piece needs memory that can keep up with a powerful processor running many programs at once: so it uses a standard called DDR, “double data rate,” which runs at a reasonably high voltage and offers high bandwidth. The processor on my iPhone is less powerful, so it needs less data at any given moment; but voltage matters enormously, since every milliwatt allocated to memory is drained from the battery. So smartphones use LPDDR, “low-power double data rate,” a variant of DDR engineered to operate at lower voltages.
The last MacBook Pro to use DDR was in 2019. All Apple Silicon Macs use LPDDR.
Apple has been using LPDDR in the MacBooks since at least 2015. I remember it was one of the complaint of the 2016-2017 MacBook Pros. They were still using LPDDR3 because LPDDR4 wasn't ready for production yet (despite regular DDR4 being available). The 2018 MacBook Pro's finally switched to LPDDR4.
But aren't there plenty of used expensive phones from the last 5-7 years that are more or less equivalent replacements for new cheap phones? Apple alone sells 250 million phones a year.
What was most surprising about all this to me was this line:
> So modern DRAM manufacturing is an extraordinarily complex and expensive process. Building a single state-of-the-art DRAM fabrication facility, a “fab,” will cost you about $15 to $20 billion; acquiring all the necessary equipment, like lithography tools and etching machines, will cost you another few billion; and then it’ll take you a few years of producing substandard and defective memory chips before your yields start to look competitive.
Extraordinarily complex and expensive! And yet I look at all the money being shuffled around between Nvidia and Google and Microsoft and Amazon and Apple and can't help but think that this is a tiny amount in comparison to what they're moving around on the stock market buying shares in each other.
Apple in particular has $20B in its couch cushions and is very vertically integrated and hardware-focused. Apple silicon is currently made by TSMC, but it seems they'd be a prime candidate to spin up their own memory fab.
I suppose the biggest problem to current executives at each company is the "few years" until that investment yields results, in the short term it's better to pay through the nose and buy GPUs with HBM at any price.
And what happens if the market settles back down or the leading memory tech pivots away from what you invested all this capital and time chasing?
You'd need a very strong, very particular forecast to make such a costly bet. And conversely, it may say something about their internal forecasts that they're not making the bet.
I’m of the (possibly unpopular) opinion that ~80% of these new data centers are never going to see the light of day.
None of the economics or supply needs of these things make any sense. The water is generally not there, electric transformers are next to impossible to acquire. I just read about one in Utah that’s supposed to be 2.5x the size of Manhattan?? https://www.theverge.com/ai-artificial-intelligence/933687/u...
I could be wrong, but I think a lot of this was planned by software guys who are used to brute-forcing their way through “impossible” problems. But those were software problems, where the limits are mostly theoretical. In the world of atoms, limitations are real.
Part of this is a lie high prices for equipment has a lot to do with monopolies in the whole supply chain and one of the major reason I am rooting for China to get parity on node as that would mean it would have been able to break all the monopolies and we get competition on the whole production from uv machines, wafer, on equipment for storing etc. Renewables are rapidly taking the world to very cheap or free energy but it would be bad if the way to best use the energy for production is controlled by a few corporations
> it seems they'd be a prime candidate to spin up their own memory fab.
While Apple et al certainly have the money to tilt up their own fab, they're savvy enough to understand the memory market's long history of constant boom/bust cycles. I still remember the huge DRAM shortage in late 80s forcing my startup at the time to delay launching our new product for a year.
People assume Apple cares about vertically integrating cost but they're actually focused on integrating margin. Apple has billions in cash on hand and when they think about what to do with it, a key metric is Return on Capital, especially the margin that capital will generate. Since a core metric public companies are judged on is blended margin, they are looking for ways their bags o' cash can be put to work generating revenue at margins that will pull their current average margin up vs down.
Averaged over time, mainstream memory devices are historically one of the worst margin areas of the semi market. It's super expensive to tilt up a fab on a new node but once you do, turning the crank faster to make a lot more chips isn't too hard because mainstream DRAM tends to be quite uniform. So when a fab on a new node and/or RAM generation first opens, the margins tend to be pretty great. But as the node matures and/or the RAM generation goes from 'new' to 'commodity', competition heats up as everyone gets better at making more faster. Then they're tempted to maximize revenue by cutting prices until their mature fab is at 101% utilization. And that eventually drives margins down until someone's selling near cost to sustain their low-price-enabling volume - with occasional dips below cost when they get stuck holding excess inventory. That's why cash-rich companies with high margins like Apple are delighted to buy DRAM built with Other People's Money. As long as the DRAM market is under competitive pressure, Apple gets to shop their huge orders around to get the absolute lowest price on RAM that was built with other investor's low margin dollars.
While I think Apple is a prime candidate to do something like this financially, I’ve never had the impression that they ever want to get involved in something “just” for vertical efficiency. There’s always a long-term vision where they can leverage it to gain a competitive edge that their competitors can’t match (the PA Semi acquisition being a perfect example of this).
RAM doesn’t seem like something where simply owning the manufacturing could lead to a disproportionate competitive edge. It would just be a vertical efficiency gamble that may or may not pay off. Of course that could simply be a failure of my imagination.
Apple, Google Microsoft Amazon should all set up memory and CPU fabs. I don’t understand why they don’t do it. $400 million ASML machine is chump change and you can almost certainly find/train locals or provide incentives to immigrants to come here as workers to fill the roles there.
> And yet I look at all the money being shuffled around between Nvidia and Google and Microsoft and Amazon and Apple
A lot of that is “weird money” created by the act of passing it around between the entities or “Holywood accounting” style money that exists when convenient and will vanish the instant it might be taxed or need to be extracted from the cycle to pay for something tangible. Trying to use a large amount of that money for tangible long-term building projects risks piercing the vail.
$20B a single Fab. And you don't just have one Fab. You need multiple. And again I have to state this every time, the most expensive Fab is empty Fab. You need to fill it, and you need to fill constant and consistently for years.
>but it seems they'd be a prime candidate to spin up their own memory fab.
They will be the prime candidate to work with memory makers for additional capacity. Not to start their own one. Here is a $20B cheque and 5 years guarantee of orders. Go and make me some memory for his price.
I mean this is the same story for iPod and NAND. It seems we are repeating what we should have learned.
For a commodity part, Apple has a history of buying a manufacturing partner a dedicated manufacturing line in exchange for locking down a guaranteed supply at a lower overall price.
It's one of the early moves Tim Cook made after joining Apple to lock down enough flash memory supply to move iPods from hard drives to flash.
I'm a newborn shill for Ulefone. They come unlocked and the manufacturer supports rooting. The devices are rugged, heavy on features and are (still) reasonably priced.
I’m really wishing I had overbuilt my NAS last year. As it stands I feel lucky to have even built it at all given I bought all the parts in the last week of September.
My 4090 and 12900k are gonna have to last till 2029 at this rate won’t they…
At Xiaomi's latest smartphone launch event, Xiaomi founder Lei Jun said that memory prices are likely to keep rising over the next two years, which could drive smartphone prices up as well. His conclusion was pretty direct: everyone should just replace their phones now. Kind of a depressing story.
This is the biggest memory repricing cycle I've ever seen in my life; some degree of high price/limited availability and "free RAM with purchase of Doritos" cycle is always expected, but this has been the worst one yet.
As other commenters have pointed out but I might have missed in the article, compute maturation is amplifying memory constraints right now and making it worse. Device upgrade cycles are getting longer because most compute-based products have matured, with CPUs not seeing substantial gains and memory usage really only expanding at the absolute top end of workloads pre-LLMs (3D and HPC in particular). An iPhone 14 still has almost all the features of the iPhone 17, because the compute capabilities are remarkably similar; Geekbench shows a performance delta of ~25-30% between the 14 and 17 Pro Max models, which is pretty paltry considering the devices are separated by four years of manufacturing improvements. This extends into desktops, laptops, tablets, STBs, and more, with only VR devices and larger ARM/RISC-based kit seeing more substantial uplifts as general designs improve.
So with compute stagnating and memory constrained, my money is on vendors taking this as an opportunity to gradually shift away from a yearly release cadence and slow down to a biennial cycle that alternates between budget and flagship launches every other year. Even if LLMs fail spectacularly and all that memory capacity becomes available, HBM memory likely isn't to find its way into many consumer devices (just ask AMD how it worked out for them on consumer GCN GPUs).
The name of the game, especially for consumers, is efficiency - "potato builds", as I've been calling them. Software and services optimized for lower power, smaller-specced devices of increasing age instead of pandering to flagship devices with poorly optimized code or engines for the sake of new shinies (like Raytracing). Between the memory shortage, shifting geopolitics, rising costs, and stagnant wages, consumer purchasing power is going to be squeezed like a vice for the foreseeable future, and businesses will need to adapt around that reality.
It's impressive that somehow, as if by coincidence, we're seeing the biggest inflationary drivers for decades, perhaps for centuries, all happening simultaneously.
The Iran war is spiking the price of oil and will likely cause shortages of pretty much everything if it isn't ended.
The Ukraine war is helping with that by destroying Russian refining capacity.
The memory shortage is set to do the same to consumer electronics, which are absolutely essential to the modern economy.
Meanwhile the AI fad is seeing huge layoffs. At the same time as the AI Big Cos are beginning to show signs of ending the subsidised free lunch phase and moving to a utility model, which will raise prices for every company that is hooked on AI.
Also tariffs. Although I'm not sure if anyone knows what's happening there.
And farms are failing. Climate change will accelerate that, so there will be food shortages within a few years.
If it's not cynical and deliberate, it's an astounding confluence of (literally) catastrophic mismanagement.
Presumably food shortages will first be felt by livestock, so expect meat to get much more affordable (as animals are slaughtered to avoid the need to feed them) for a while, then expensive (when there are fewer head). Animal feed represents over 50% of acreage.
Maybe it's time to go back to the time where computers and entertainment were niche and develop efficient resource-constrained devices....
I hate it that we had decades of progress to have computers become a very expensive hobby because some dudes high on fentanyl think some text prediction model that destroys the planet is worth a trillion dollars.
I've been buying a "new" used iphone for $100-150 now, every few years, for over a decade. My "new" used Dell laptop I bought a few months ago for $40 which became a linux system in just an hour. All good here.
I appreciated the detailed breakdown of the memory crunch and how it will affect parts of the industry and consumers. Very good article.
I'm not one of those people who chases all the new great things. I wait until things wear out or become completely obsolete before upgrading. I just get comfortable doing things the same way every day and see no reason to waste money on SaaS shit or anything else wastes my time or money.
I think the memory shortage will present opportunities for those willing to take advantage of the situation. A lot of DRAM is going into GPUs for data centers in AI work. Those units have a limited lifetime online and they will be rotated out and replaced with new units as performance degrades. I think this will be a lot like Li-ion batteries in that many of these GPUs will be perfectly fine for home pcs or small business workstations or for other less intensive use cases and the RAM will be performant enough that a viable recycling industry should arise from this AI buildout.
Funny enough, one day the local AI noise-making, power-wasting, water-wasting data centers will be the best places to score high-tech components and many of us will have one right down the road. That should set a lot of people up as recyclers redistributing reconditioned components to those who build their own systems.
I heard that one company buying all the product or a majority of the product is called a monopoly practice.
If a dominant buyer locks up most of the supply chain through exclusive contracts, it prevents rival companies from getting the materials they need to survive, which violates laws like Section 2 of the Sherman Antitrust Act.
This is another place where modern capitalism struggles in an area that requires large capital investment producing payoff after years of startup costs, and the buy side is volatile. We've seen this recently with rare earths, copper, and some other minerals. DRAM is almost in the same category.
China has an advantage here, once domestic DRAM production finally gets going.
DRAM policy can be set strategically. China's economic planners may choose to provide DRAM to domestic manufacturers rather than export parts, even if exporting parts would be more profitable in the near term. That's already being done in raw materials.
Conversely, if external suppliers have lower prices, there may be a policy decision to buy domestically to keep the domestic manufacturers going. Done with the goal of leveling production, this can work. Done stupidly, it becomes a money drain, of course.
Probably China controls the DRAM market around 2030 or so.
I work in healthcare imaging. The DRAM crunch is translating into significant increases in costs - because reconstructing images from raw scans is a compute-itensive process; and that gradually translates into product price increases. And then, healthcare providers may well opt for cheaper, lower-tier medical scanners, since they can't afford the better ones; and then finally you get less accurate scans if you're suspected to have cancer or whatever.
So, that's another way we are financing the LLM machine and the trillion-dollar valuations of those corporate behemoths.
So, if you were wondering how come LLM services are cheap or gratis - now you know another group of people bearing the costs: The poorer half of humanity, who will now not be able to afford a smartphone, i.e. not having access to a computer. Sure, that doesn't kill you, but it is quite significant personally and socially.
So the availability of cheap phones is going down because of the cost of RAM.
What about the RAM consumption trend of the last 10 years? I think it is very feasible to produce phones with the same amount of RAM as was the norm 10 years ago. The only compromise would be using older algorithms and features that consume less of it, and to take a bit of effort on keeping an eye on memory consumption in the development phase. There's a lot of opportunity. We can even leverage AI these days to optimize existing software for RAM usage.
90 comments
[ 2.3 ms ] story [ 115 ms ] thread> these memory makers have learned a very particular lesson from the unforgiving history [deep drops in demand] of their industry: always leave demand unmet
I can't blame them for keeping some reserve demand ready so they keep having customers over the years.
From 2008 to around 2015, upgrading every two years could make a meaningful difference. From ~2015 to ~2020 upgrading every three years might be worth considering. I just upgraded my top of the line flagship after nearly six years. And I actually looked for compelling reasons to buy a new phone every year since 2023. There just weren't any.
Frankly, this latest flagship phone is pretty underwhelming. It's slightly faster at a few things. The battery lasts a little longer. The screen can get a little brighter. The camera is supposed to a little better. But those are just the claimed improvements. I haven't actually noticed any of them in daily use because they weren't issues with my 2020 flagship phone either. Otherwise, the new phone is almost exactly the same size, same weight, same resolution, same look and same capabilities. I only upgraded because I was long out of contract and it was a only a couple hundred bucks for a $1400 MSRP phone with a new contract and a trade-in of the old phone.
I was using an iPad Pro from late 2018 (mostly just for casual web browsing, reading documents and watching video on, I still do all my real work on laptop/desktops) as well until this year, and would have kept using it if I hadn't accidentally dropped it in water. I don't really notice much difference at all between my old one (when it worked) and the new iPad Air I replaced it with, except for the battery being a little better and having a bit more ram being nice (websites in background tabs are less likely to be purged from memory when I come back to them).
For anyone who doesn’t follow the market closely, this is about a good a primer as you could hope for.
> So modern DRAM manufacturing is an extraordinarily complex and expensive process. Building a single state-of-the-art DRAM fabrication facility, a “fab,” will cost you about $15 to $20 billion; acquiring all the necessary equipment, like lithography tools and etching machines, will cost you another few billion; and then it’ll take you a few years of producing substandard and defective memory chips before your yields start to look competitive.
Extraordinarily complex and expensive! And yet I look at all the money being shuffled around between Nvidia and Google and Microsoft and Amazon and Apple and can't help but think that this is a tiny amount in comparison to what they're moving around on the stock market buying shares in each other.
Apple in particular has $20B in its couch cushions and is very vertically integrated and hardware-focused. Apple silicon is currently made by TSMC, but it seems they'd be a prime candidate to spin up their own memory fab.
I suppose the biggest problem to current executives at each company is the "few years" until that investment yields results, in the short term it's better to pay through the nose and buy GPUs with HBM at any price.
You'd need a very strong, very particular forecast to make such a costly bet. And conversely, it may say something about their internal forecasts that they're not making the bet.
None of the economics or supply needs of these things make any sense. The water is generally not there, electric transformers are next to impossible to acquire. I just read about one in Utah that’s supposed to be 2.5x the size of Manhattan?? https://www.theverge.com/ai-artificial-intelligence/933687/u...
I could be wrong, but I think a lot of this was planned by software guys who are used to brute-forcing their way through “impossible” problems. But those were software problems, where the limits are mostly theoretical. In the world of atoms, limitations are real.
The amounts of money circulating whilst some of us struggle to make rent ...
Nothing fair, or just, about this world we live in
I wonder how long it REALLY took them to move from intel to apple silicon, which they don't even make.
It might be easy, like a consumer deciding to generate their own electricity (pv on the roof)
or it might be slightly harder, like a consumer deciding to generate their own electricity by drilling for oil, refining it, etc...
While Apple et al certainly have the money to tilt up their own fab, they're savvy enough to understand the memory market's long history of constant boom/bust cycles. I still remember the huge DRAM shortage in late 80s forcing my startup at the time to delay launching our new product for a year.
People assume Apple cares about vertically integrating cost but they're actually focused on integrating margin. Apple has billions in cash on hand and when they think about what to do with it, a key metric is Return on Capital, especially the margin that capital will generate. Since a core metric public companies are judged on is blended margin, they are looking for ways their bags o' cash can be put to work generating revenue at margins that will pull their current average margin up vs down.
Averaged over time, mainstream memory devices are historically one of the worst margin areas of the semi market. It's super expensive to tilt up a fab on a new node but once you do, turning the crank faster to make a lot more chips isn't too hard because mainstream DRAM tends to be quite uniform. So when a fab on a new node and/or RAM generation first opens, the margins tend to be pretty great. But as the node matures and/or the RAM generation goes from 'new' to 'commodity', competition heats up as everyone gets better at making more faster. Then they're tempted to maximize revenue by cutting prices until their mature fab is at 101% utilization. And that eventually drives margins down until someone's selling near cost to sustain their low-price-enabling volume - with occasional dips below cost when they get stuck holding excess inventory. That's why cash-rich companies with high margins like Apple are delighted to buy DRAM built with Other People's Money. As long as the DRAM market is under competitive pressure, Apple gets to shop their huge orders around to get the absolute lowest price on RAM that was built with other investor's low margin dollars.
RAM doesn’t seem like something where simply owning the manufacturing could lead to a disproportionate competitive edge. It would just be a vertical efficiency gamble that may or may not pay off. Of course that could simply be a failure of my imagination.
Even if you were nowhere near state of the art, being able to produce millions of your own cards every year at cost would save you a lot of money.
Nvidia sells all the stock, prices for GPUs moved up not down.
Companies like Apple don't want to be in every industry. Its risk, its cost, its expertise you need.
And sure you can throw money at it but you need people. Alone hiring all the experts is probalby not that easy. Who knows how to setup a DRam Factory?
It takes years to build this up
A lot of that is “weird money” created by the act of passing it around between the entities or “Holywood accounting” style money that exists when convenient and will vanish the instant it might be taxed or need to be extracted from the cycle to pay for something tangible. Trying to use a large amount of that money for tangible long-term building projects risks piercing the vail.
Huge gamble - If they pull it off I wouldn't be surprised if other companies follow
>but it seems they'd be a prime candidate to spin up their own memory fab.
They will be the prime candidate to work with memory makers for additional capacity. Not to start their own one. Here is a $20B cheque and 5 years guarantee of orders. Go and make me some memory for his price.
I mean this is the same story for iPod and NAND. It seems we are repeating what we should have learned.
It's one of the early moves Tim Cook made after joining Apple to lock down enough flash memory supply to move iPods from hard drives to flash.
Pics:https://duckduckgo.com/?ia=images&origin=funnel_home_website...
Two other underappreciated handset brands are Doogee and Blackview. Gorgeous devices and solidly built. From what I recall they're friendly to root.
My 4090 and 12900k are gonna have to last till 2029 at this rate won’t they…
As other commenters have pointed out but I might have missed in the article, compute maturation is amplifying memory constraints right now and making it worse. Device upgrade cycles are getting longer because most compute-based products have matured, with CPUs not seeing substantial gains and memory usage really only expanding at the absolute top end of workloads pre-LLMs (3D and HPC in particular). An iPhone 14 still has almost all the features of the iPhone 17, because the compute capabilities are remarkably similar; Geekbench shows a performance delta of ~25-30% between the 14 and 17 Pro Max models, which is pretty paltry considering the devices are separated by four years of manufacturing improvements. This extends into desktops, laptops, tablets, STBs, and more, with only VR devices and larger ARM/RISC-based kit seeing more substantial uplifts as general designs improve.
So with compute stagnating and memory constrained, my money is on vendors taking this as an opportunity to gradually shift away from a yearly release cadence and slow down to a biennial cycle that alternates between budget and flagship launches every other year. Even if LLMs fail spectacularly and all that memory capacity becomes available, HBM memory likely isn't to find its way into many consumer devices (just ask AMD how it worked out for them on consumer GCN GPUs).
The name of the game, especially for consumers, is efficiency - "potato builds", as I've been calling them. Software and services optimized for lower power, smaller-specced devices of increasing age instead of pandering to flagship devices with poorly optimized code or engines for the sake of new shinies (like Raytracing). Between the memory shortage, shifting geopolitics, rising costs, and stagnant wages, consumer purchasing power is going to be squeezed like a vice for the foreseeable future, and businesses will need to adapt around that reality.
The Iran war is spiking the price of oil and will likely cause shortages of pretty much everything if it isn't ended.
The Ukraine war is helping with that by destroying Russian refining capacity.
The memory shortage is set to do the same to consumer electronics, which are absolutely essential to the modern economy.
Meanwhile the AI fad is seeing huge layoffs. At the same time as the AI Big Cos are beginning to show signs of ending the subsidised free lunch phase and moving to a utility model, which will raise prices for every company that is hooked on AI.
Also tariffs. Although I'm not sure if anyone knows what's happening there.
And farms are failing. Climate change will accelerate that, so there will be food shortages within a few years.
If it's not cynical and deliberate, it's an astounding confluence of (literally) catastrophic mismanagement.
I hate it that we had decades of progress to have computers become a very expensive hobby because some dudes high on fentanyl think some text prediction model that destroys the planet is worth a trillion dollars.
I'm not one of those people who chases all the new great things. I wait until things wear out or become completely obsolete before upgrading. I just get comfortable doing things the same way every day and see no reason to waste money on SaaS shit or anything else wastes my time or money.
I think the memory shortage will present opportunities for those willing to take advantage of the situation. A lot of DRAM is going into GPUs for data centers in AI work. Those units have a limited lifetime online and they will be rotated out and replaced with new units as performance degrades. I think this will be a lot like Li-ion batteries in that many of these GPUs will be perfectly fine for home pcs or small business workstations or for other less intensive use cases and the RAM will be performant enough that a viable recycling industry should arise from this AI buildout.
Funny enough, one day the local AI noise-making, power-wasting, water-wasting data centers will be the best places to score high-tech components and many of us will have one right down the road. That should set a lot of people up as recyclers redistributing reconditioned components to those who build their own systems.
China has an advantage here, once domestic DRAM production finally gets going. DRAM policy can be set strategically. China's economic planners may choose to provide DRAM to domestic manufacturers rather than export parts, even if exporting parts would be more profitable in the near term. That's already being done in raw materials. Conversely, if external suppliers have lower prices, there may be a policy decision to buy domestically to keep the domestic manufacturers going. Done with the goal of leveling production, this can work. Done stupidly, it becomes a money drain, of course.
Probably China controls the DRAM market around 2030 or so.
my 2 cents
So, that's another way we are financing the LLM machine and the trillion-dollar valuations of those corporate behemoths.
What about the RAM consumption trend of the last 10 years? I think it is very feasible to produce phones with the same amount of RAM as was the norm 10 years ago. The only compromise would be using older algorithms and features that consume less of it, and to take a bit of effort on keeping an eye on memory consumption in the development phase. There's a lot of opportunity. We can even leverage AI these days to optimize existing software for RAM usage.